% notes from discussion, at Panel on Semantic Web Services % at WWW-2003 Conference % notes by Benjamin Grosof (panel chair) http://ebusiness.mit.edu/bgrosof , % with help/input from Michael Dean, 5/23/2003 % other panel materials (slides, handouts) are posted at Benjamin's webpage little poll at the beginning: about half know what Semantic Web Services means, among panel and audience Benjamin prepared remarks: Q: relationship of DAML-S to SWSI? his A: DAML-S is sort of molting into SWSI which picks up from DAML-S Christoph prepared remarks: some highlights www.zapthink.com -- attempts to classify XML stds/spec's don't ignore 30 years of experience with EDI (e.g. unwillingness to expose internal processes) killer app: B2B integration e.g., 300K co's use EDI today challenge: make a new pair of co's work together in 5min Avi: highlights of prepared remarks: History - learn it or repeat it Simplicity - KISS Avi's mother hasn't learned HTML yet one person using (LC)lint relev prev wk: - Curtis et al '92: many aspects to a process/service: functional, behavioral, i/o info, org'al - '77 Michael Hammer: very h-l prog lang for integr - PIF/PSL: process interchange (uses KIF) want to make the SW active Bijan: highlights of prepared remarks: obstacle: lack of understanding of the web real power is when you publish something on your own web site danger: UDDI will eventually get around to reinventing what we're doing with the Semantic Web - and vice versa attraction: principled ad hoc'ery very late, situationally directed composition of human and program activity Sheila in absentia: highlights of prepared remarks [presented by Benjamin]: obstacle: need for widespread adoption opportunity intra-enterprise from trust being less problematic %%% General Discussion: Q: how do we provide biz case? Avi A: incentive problem, rather than technical problem - different for different industries and situations (e.g. increasing Google hit rates) e.g., go to big Swiss banks who have the data, then step in and help -- requires understanding their problems; they have been working on integration for years some biz cases: 24x7 / reliability, provable correctness, adapting/flexib Q [by Roger Cutler of ChevronTexaco]: there's no lack of biz case to do data integr, incl. to provide metadata for data which is now mostly unusabble. But most such folks have never heard of us, they talk about taxonomies -- which part of this technol do we want to make visible Chris A: yup, don't know what it is Q by Jim Hendler: is this a disruptive technology, in a good sense? -- if not, why do it? -- if yes, then in what ways? Bijan A: - new app's that weren't obvious, as well as cost reductions Chris A: yes, it's disruptive to enable a biz proc expert in my org to set up a new B2B partner relnsh in *5 minutes* Q: who can do it in 5min when will a biz person intra-org be able to do it, using own terms A: lots of work Q: ad hoc integration in distrib heterog env is same as agent rhetoric - is there learning from that, place for that in this work? A: yes, from both success and mistakes Q (ff-up): what are the successes and mistakes? A: doesn't seem well worked out: - how web works - how you apply other things to the web - nobody understanding it Q: what are the benefits/lessons of agents by Nick Gibbins: same rhetoric that agent community has been using for years? are you leveraging their experience? Bijan: no convincing theory about how the web works Benjamin A: some lessons from experience with agents: 0. agent is a perspective, there are multiple defn's -- I define an agent as a knowledge-based application, thus uses/overlaps-highly SW/SWS 1. Knowledge-based part is usually only a part of a successful app; humility 2. Don't let the term (e.g., agents or SWS) be so vague it's hijackable 3. there is a variety of useful knowledge-based techniques / theories: incl. data mining / machine learning, Bayesian decision analysis, rules, ontologies, game theory Q: is it too early for SWS, given that WS is still struggling with security and transactions Bijan A: no; otherwise will probably have to go back and fix WS things; can co-evolve, both sides organically moving together Benj A: WS needs SW to succeed Bijan more A [to a follow-up Q]: WS doc's return again and again to "we need semantics" Q by James Faruggia: how to make SWS efforts visible, if govt funding stopped? Chris A: biz's are asking for it, even aside from funding -- [they recognize] integration is important Q: how does this relate to movement towards model-based deployment, e.g., UML being used to configure compon's and workflows, w/ human in the loop Chris A: workflow often model-driven, then needs human in the loop, but can have ontol model present even before the human gets into the loop - so is a prerequisite Q: what are you doing to help SWS happen? Christoph A: being here [at the panel / conference] Bijan A: - on SWSI Language - doing end to end system Avi A: ..., MIT Process Handbook repository of biz processes - doing a couple efforts now: . SWS discovery and matchmaking . export PH as an ontology (w/ Benjamin) Chris A: - in SWSI Architecture - ... workflow work... - architecting B2B integration product at Oracle Q: interactions with other stds groups? Bijan A: W3C WSDL and WS-Arch they like DAML-S Chris A: rep to influence all the stds in integration Avi A: big s/w co's that want to include biz proc aspects Benj A: besides app scenarios: rules/KR for SWS Oasis, incl. Legal XML, eContracts, policy; education/linkage about SWS in conversation with individual co's, e.g., participants in RuleML Initiative Bijan Q to Chris: aren't we/SWSI disconnected to W3C e.g., does Arch talk to WS-A Chris A: yes, is connected: we put all technical discussion of SWSI on w3c www-ws mailing list Q: will SWSI be in contact w/ Grid efforts A: we have a rep/liaison; we plan to make sure comment: there's innovation diffusion usually 17yrs to saturation in usual sigmoid, maybe 2X faster now; but probably decade timeline is approp. A: Chris: sometimes faster in that there are pockets that are innovative and you establish semantics in a communal agreement way not a formal way; thus should aim to identify Q: when will web services be self-introspective? Christoph A: 11 years Q: will SWS influence WS? Bijan A: yes for some easy wins (e.g using ontologies for taxonomies) Chistoph A: www.amazon.com/webservices/ Bijan A: beware Web orthodoxy, i.e., sometimes need to rethink more radically, doing something new Chris A: trend: integration into data manipulation, app server and OS world, trickle deeper into the general IT infrastructure; is a very positive sign JimH [Hendler] comment: wrt: is it too soon to do SWS?: the first of the real WS's are coming on line -- we need unify these or will be 17yr as people have to retrofit - if it's high value/work ratio, will become accepted soon - i.e., help make WS's work correctly filtering, integration, discovery Q: will UDDI have rich data models Bijan A: yes, insofar can do with a small move -- e.g., stick in an ontology where they have a taxonomy -- but they're fighting the Web spirit of publishing control Q: to what extent does WS really need SWS to succeed? Benj A: to achieve the fuller WS vision incl high degree of automation of support to Chris A: Amazon currently: ex. of doing useful WS wo/ SWS Avi A: often/usually those WS's have IMPLICIT semantics - argument of SWS is that there's lots of leverage out of making Comment: only 1% of world's co's do EDI even for part - partly because it's hard -- incl. to understand the biz processes themselves before you can automate [them in a (S)WS'y way] Avi response: incrementally we'll keep lots of human activity unautomated - that's the nature of economics/society Comment/Q: there's lots of tools for WS's, what about tool vendors in SWS space Benj A: SWS as fused area is in quite-early/mainly-research days, tho' there are a few early tools e.g. from DAML-S; WS builders can use many existing SW tools however, e.g., RDF, data integration, ontologies, rules, interoperability JimH: there are four startups/spinoffs in the space; but biz model is typically to use own tools as compet adv, then eventually commercialize the toolsets, e.g., how they do data things B2B is main playground B2C wide open small-co integration is wide open -- e.g., says Gartner -- and yet is mainly about extra-enterprise Bijan: Semantic Web Services are part of the aftermarket for RDF and OWL Comment: [Chevron person: Roger Cutler] on potential early adoption areas: opportunity for equivalent of EDI stuff to SME's, e.g., purchasing EDI, but with lower cost of entry which inhibits small-co's (can't afford million-buck middleware) - so SUGGESTION: go and find out how EDI works Chris response: e.g., Boeing has 30,000 suppliers some of which are tiny (e.g., 8 persons for producing lights ?on seats?) Benj response: big co's w/ large supplier bases -- e.g. Boeing, TRW and automotive, where only like 10% use EDI -- can be a key catalyst, tho' - they have strong incentive and market power to impose top-down the adoption of a new technol even by smaller co's - so SWS could take off that way once reaches a cost-effectiveness threshold Q: lots of sites that isn't well structured or accessible where principles of SW technol can be applied, to show real benefit wo/ needing a "big bang"; look for where it's very visible, e.g., customer-facing Chris response: EDI mkt is growing still, because the Internet bubble made the benefits of automated integration more visible, then they found EDI/Swift/established environments - is evidence that when Avi response: emergency response: firefighters, police, ... - all have info services that would like to do exchange, but they're very incompatible - whole different sector, where the tradeoff isn't nec'ly to be cheaper, but to do new things Benj: yes, there the objective function isn't nec'ly money